CN105956010B - Distributed information retrieval set option method based on distributed characterization and partial ordering - Google Patents
Distributed information retrieval set option method based on distributed characterization and partial ordering Download PDFInfo
- Publication number
- CN105956010B CN105956010B CN201610251677.1A CN201610251677A CN105956010B CN 105956010 B CN105956010 B CN 105956010B CN 201610251677 A CN201610251677 A CN 201610251677A CN 105956010 B CN105956010 B CN 105956010B
- Authority
- CN
- China
- Prior art keywords
- distributed
- document
- scoring
- sample
- word
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000012512 characterization method Methods 0.000 title claims abstract description 34
- 239000013598 vector Substances 0.000 claims abstract description 42
- 238000005070 sampling Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 description 6
- 238000012545 processing Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of distributed information retrieval set option methods based on distributed characterization and partial ordering, it include: to receive original query from the user, it is extended the inquiry that is expanded to original query, and calculates the distributed characterization vector of the expanding query；For any one document in the sample set of each set, the distributed characterization vector of the document is calculated, and using the cosine value of the angle between the document distributed characterization vector corresponding with expanding query as the scoring of the document；For any one set, the scoring of the set is calculated according to the scoring of document each in the sample set of the set, and the higher k set that select to score is as final result；The sample set of each set is by obtaining the pooled sampling.The present invention indicates document and inquiry using distributed characterization vector, and using inquiry and set relatedness computation based on partial ordering, introduces document scores threshold value, improve the accuracy of set scoring, and then improve retrieval accuracy.
Description
Technical field
The present invention relates to distributed information retrieval technical fields, and in particular to one kind is based on distributed characterization and partial ordering
Distributed information retrieval set option method.
Background technique
Distributed information retrieval (Distributed Information Retrieval, DIR) system is usually by big document
Collection (number of documents is more) is divided into several small document sets (number of documents is few), and each small document sets (referred to as gathering) are serviced by one
Device separate storage and retrieval.After receiving the inquiry of user, distributed information retrieval system will be inquired while is transmitted to multiple
Server, then the result that server returns is merged, it finally returns to user.Under normal circumstances, it inquires and each set
The degree of correlation is different, and retrieves expense to reduce, and distributed information retrieval system usually first calculates the degree of correlation of inquiry with set,
Obtain set scoring；It again will set descending arrangement by set scoring；Finally forward the query to k set place in the top
Server, this process is referred to as Resource selection.
In the latest 20 years, many researchs have been emerged in large numbers in distributed information retrieval Resource selection field.A portion method will
Set is considered as one " super large document " (big document).Such as CVV (The Cue-Validity-Variance) and CORI
The methods of (Collection Retrieval Information Network) uses the system such as dictionary, word frequency rate and document frequency
The scoring of information set of computations is counted, these methods not only have ignored set sizes, also require each set offer that can describe its own
Dictionary and the statistical informations such as word frequency, this is difficult to realize under miscoordination formula environment.
Another part method, which regards set as, to be made of numerous small documents.Such as ReDDE (Relevant
Document Distribution Estimation)、CRCS(Central-rank-based Collection
Selection it) is used with the methods of SHIRE (Sampling-based Hierarchical Relevance Estimation)
The keyword relevance of TF-IDF form and the fitting function of predicting function calculate the degree of correlation of inquiry and document (for statement
Convenient, " degree of correlation of inquiry and document " is referred to as " scoring of document " by the present invention), but have ignored semantic information.
Matthias et al. indicates inquiry and set using ESA vector, and scores vector similarity as set, however ESA vector
It is influenced vulnerable to dimension calamity.In addition, the document ordering mode of existing set option method is also unreasonable, and under normal circumstances, Yong Hugeng
It is concerned about and inquires maximally related document, therefore gathering scoring should be directly proportional to the degree of correlation of most relevant documentation, and existing document is arranged
Sequential mode can then miss the most relevant documentation of partial set.
Summary of the invention
In view of the deficiencies of the prior art, believed the present invention provides a kind of based on distributed characterization and the distributed of partial ordering
Breath retrieval set option method, this method have both recall precision height and the high advantage of retrieval accuracy.
A kind of distributed information retrieval set option method based on distributed characterization and partial ordering, comprising:
Step 1, original query from the user is received, the inquiry that is expanded is extended to original query, and calculating should
The distributed characterization vector of expanding query；
Step 2, for any one document in the sample set of each set, calculate the document distribution characterize to
It measures, and characterizes cosine value the commenting as the document of the angle between vector using the document distribution corresponding with expanding query
Point；
Step 3, for any one set, which is calculated according to the scoring of document each in the sample set of the set
Scoring, and the higher k set that select to score is as final result；
The sample set of each set is by obtaining the pooled sampling.
Using combining the enquiry expanding method of Wikipedia and ListNet to be extended original query, detailed process is such as
Under:
Step 100, it is retrieved in all webpages of Wikipedia according to the keyword of original query, retrieval is obtained
Web page title as candidate expansion word；
Step 101, for each candidate expansion word, existed according to the keyword of candidate's expansion word and original query
The case where abstract and body part of each webpage of Wikipedia occur calculates the feature vector of candidate's expansion word, and calculates
Scoring of the inner product of this feature vector and weight vectors as candidate expansion word,
Preferably, described eigenvector is identical with the dimension of weight vectors, the weight vectors use ListNet algorithm
Training obtains；
Step 102, will score several higher (specific number can need to set according to application) candidate expansion word conducts
Keyword increases to original query and obtains expanding query.
The distributed characterization vector of expanding query is calculated according to the following formula in the present invention:
Wherein, Vq'For the distributed characterization vector of expanding query q', VtermThe distribution of word term is obtained for what is precalculated
Formula characterizes vector, tftermThe word frequency rate for being word term in expanding query q'.
Further preferably, the distributed characterization vector of each document and word term pass through PV model training and obtain.
Preferably, the sample set of each set is by using the sampling based on inquiry according to default sample rate to the set
Method samples to obtain.
Further preferably, the scoring of step 3 set of computations c includes:
Step 300, it determines that the document for meeting following condition is used as most relevant documentation from the sample set of the set, and is formed
Most set of relevant documents:
Wherein, dlFor in sample set score descending sort when ranking be l document,For sample dlScoring, τcFor
For the preset scoring threshold value of set c, ωcFor for the preset scoring rank threshold of set c；
Step 301, the scoring R of set of computations c according to the following formulac:
Wherein, αcSample rate when to be sampled to set c, πcFor the most set of relevant documents of set c.
To guarantee to inquire precision, τ in the present inventioncIt sets according to the following formula:
τc=β × Dd,
Wherein, DdFor the scoring for the highest document d that scores in sample set, β is global parameter, and value range is [0,1].
In the present invention, k, l, ωc、β、αcIt needs to be set according to practical application request in practical application.
Compared with prior art, the present invention has the advantage that
1) present invention indicates document and inquiry using distributed characterization vector, and is obtained using neural network language model
Distribution characterization vector, improves the accuracy of document semantic acquisition, to improve inquiry and the accuracy of file correlation；
2) original query is extended using the enquiry expanding method of combination Wikipedia and ListNet.Pass through introducing
Wikipedia improves the quality of expansion word, while introducing word frequency rate, document frequency and three category feature of Term co-occurrence and based on spy
The study sort algorithm ListNet of sign improves the accuracy of query semantics acquisition；
3) using inquiry and set relatedness computation method based on partial ordering, in the sortord that redefines document and
On the basis of weight calculation mode, document scores threshold value is introduced, improves the accuracy of set scoring, further increases retrieval essence
Exactness.
Detailed description of the invention
Fig. 1 is the information retrieval set option method flow chart based on distributed characterization and partial ordering of the present embodiment；
Fig. 2 is the sampling algorithm flow chart based on inquiry；
Fig. 3 is to calculate document scores sub-stage flow chart；
Fig. 4 is the flow chart of choice set zygote stage.
Specific embodiment
Below in conjunction with specific the drawings and specific embodiments, the present invention is described in detail.
The invention proposes the distributed information retrieval set option method based on distributed characterization and partial ordering, the party
Method usesIndicate a set, NcIndicate the number of document in set c, a distributed information
Retrieval environment includes multiple set { c1,c2,…,cM, M is set number.scIndicate the sample set of set c,Indicate the central sample collection of distributed information retrieval system.To improve recall precision, in advance
The some general amounts that can be calculated in advance or general procedure are used for subsequent query.
Information retrieval set option method flow chart such as Fig. 1 institute based on distributed characterization and partial ordering of the present embodiment
Show, is divided into two stages of pretreatment and online processing.
Pretreatment stage
It is pretreated that specific step is as follows:
1) using " sampling based on inquiry " algorithm is that set c constructs sample set sc, process is as shown in Figure 2.First from
1 word is randomly selected in inquiry log as initial query word；Then in the retrieval of every wheel, preceding 5 documents of return are added
To sample set sc, then from scIn randomly select the query word that 1 word is retrieved as next round；Work as scWhen middle number of documents reaches 400
Stop sampling；Sample set to all set, which constructs, to be completed, and central sample collection S can be obtained；
2) central sample collection S is input to PV model to be trained, obtains central sample and concentrates the corresponding distribution of document
Characterize vector VdDistributed characterization vector V corresponding with word each in documenttermEtc. information；
3) Wikipedia is parsed, the corpus statistics information such as the word occurred in Wikipedia and corresponding TF and DF are obtained
(being used for construction feature vector)；
4) index is established for all webpages of Wikipedia, (is made at " calculate document scores " with providing search function
With)；
5) weight vectors w (being used in the step 3 of " calculating document scores ") is trained using ListNet algorithm.
The input of ListNet is a group polling Q={ q1,q2,q3,…,qm, each inquiry qiAll correspond to a word listOne degree of correlation scoring listWith a feature vector
ListWherein niIndicate eiMiddle element number；It indicates and inquiry qiThe degree of correlation be
Word,Indicate wordCorresponding feature vector；Title from Wikipedia；siIn element be descending arrangement,
And it is calculated according to formula (1)
Wherein σ (*) is performance measure function, and performance measure function, accuracy are measured using accuracy in the present embodiment
P@n calculation formula is as follows,
Num indicates the number of documents associated with the query in the preceding n document that retrieval obtains, and n takes 10 in the present embodiment.
The online processing stage
The online processing stage, which is divided into, calculates document scores and selection two sub-stages of set.
The process for calculating document scores sub-stage is as shown in Figure 3, the specific steps are as follows:
1) enabling q is the original query of user's input, all webpages of Wikipedia (each webpage is considered as a document)
Text in retrieved, keyword q, the title of the webpage retrieved is as candidate expansion word；
It 2) is candidate expansion word e construction feature vector f (e)=[f1(e),f2(e),…,f12(e)].Formula (3) to (8)
For the feature in abstract fields, wherein formula (3) and (4) are word frequency rate TF features, and formula (5) and (6) are document frequency DF special
Sign, formula (7) and (8) are co-occurrence co-occurrence features；tf(e|fieldabstract) indicate in all of Wikipedia
In the abstract fields of webpage, the number of word e appearance；df(e|fieldabstract) indicate in all webpages of Wikipedia, it plucks
Wanting field includes the webpage number of word e；tkFor the word in inquiry q, C (tk,e|fieldabstract) indicate Wikipedia institute
Have in webpage, abstract fields include simultaneously word tkWith the webpage number of e；(tk,tr) indicate by any two query word structure in q
At word to (unordered), θ is by all words of q to the set constituted, | θ | be the element number of set θ, C (tk,tr,e|
fieldabstract) indicate in all webpages of Wikipedia, abstract fields include simultaneously word tk、trWith the webpage number of e；
Feature (i.e. f in body field2(e), f4(e), f6(e), f8(e), f10(e) and f12(e)) similar with abstract fields；
3) the scoring z of candidate expansion word e is calculated using formula (9)e, wherein " " represents the inner product of two vectors, and press
According to scoring zeAll candidate expansion word descendings are arranged；
ze=f (e) w (9)
4) it chooses γ word in the top to be appended in inquiry q, is expanded and inquires q '；
5) the corresponding distributed characterization vector V of q ' is calculated according to formula (10)q′, wherein tftermIt is word term in q '
Word frequency rate；
Vq′=∑term∈q′Vterm×tfterm, (10)
The word frequency rate of each word of the present embodiment is the ratio of word sum in the number that the word occurs in queries and inquiry.Example
Such as: inquiry is " today, weather was excellent ", including " today ", " weather ", "true" and " stick " 5 words, and the word frequency rate of this 5 words is divided
It is not exactly 1/6,2/6,1/6,1/6 and 1/6.
6) V is calculated according to formula (11)q′With VdBetween vector cosine value, and as the scoring D of document dd；
Dd=cos (Vq′,Vd) (11)
7) step 6) is repeated, is finished up to the scoring of all documents in S calculates.
The process of choice set zygote stage is as shown in Figure 4, the specific steps are as follows:
1) by the sample set s of set ccIn all documents according to document scores descending arrange, obtain document sequence
2) according to the document scores threshold tau of formula (12) set of computations cc, wherein d is scMiddle document scores DdMaximum text
Shelves, β are the global parameters that a value range is [0,1]；
τc=β × Dd (12)
3) all most relevant documentations of set c are found out.Enable dlIt is document sequenceIn a document, l dl
Ranking, ωcIt is a positive integer, in sample set scAll documents in, meet formula (13) document be exactly set c most
Relevant documentation, and use πcIndicate all most relevant documentations of set c；
4) according to the scoring R of formula (14) set of computations cc, wherein αcFor the decimation factor of set c, the i.e. document of set c
Total number and its sample set scDocument total number ratio, g (l) is the weighting function of document；
5) step 1 is repeated to 4, until the scoring of all set is calculated and finished；
6) all set are chosen into k set in the top according to set scoring descending arrangement.
Technical solution of the present invention and beneficial effect is described in detail in above-described specific embodiment, Ying Li
Solution is not intended to restrict the invention the foregoing is merely presently most preferred embodiment of the invention, all in principle model of the invention
Interior done any modification, supplementary, and equivalent replacement etc. are enclosed, should all be included in the protection scope of the present invention.
Claims (7)
1. a kind of distributed information retrieval set option method based on distributed characterization and partial ordering, which is characterized in that packet
It includes:
Step 1, original query from the user is received, the inquiry that is expanded is extended to original query, and calculates the extension
The distributed characterization vector of inquiry；The distributed characterization vector of the expanding query is calculated according to the following formula:
Wherein, Vq'For the distributed characterization vector of expanding query q', VtermThe distributed table of word term is obtained for what is precalculated
Levy vector, tftermThe word frequency rate for being word term in expanding query q'；
Step 2, for any one document in the sample set of each set, the distributed characterization vector of the document is calculated, and
Using the cosine value of the angle between the document distributed characterization vector corresponding with expanding query as the scoring of the document；
Step 3, for any one set, commenting for the set is calculated according to the scoring of document each in the sample set of the set
Point, and the higher k set that select to score is as final result；
The sample set of each set is by obtaining the pooled sampling.
2. the distributed information retrieval set option method as described in claim 1 based on distributed characterization and partial ordering,
It is characterized in that, being extended using the enquiry expanding method in conjunction with Wikipedia and ListNet to original query.
3. the distributed information retrieval set option method as claimed in claim 2 based on distributed characterization and partial ordering,
It is characterized in that, being extended in step 1 to original query and including:
Step 100, it is retrieved in all webpages of Wikipedia according to the keyword of original query, the net that retrieval is obtained
Page head is as candidate expansion word；
Step 101, for each candidate expansion word, existed according to the keyword of candidate's expansion word and original query
The case where abstract and body part of each webpage of Wikipedia occur calculates the feature vector of candidate's expansion word, and calculates
Scoring of the inner product of this feature vector and weight vectors as candidate expansion word,
Described eigenvector is identical with the dimension of weight vectors, and the weight vectors are obtained using the training of ListNet algorithm；
Step 102, several higher candidate expansion words will be scored as keyword increase to original query to be expanded and look into
It askes.
4. the distributed information retrieval set option method as described in claim 1 based on distributed characterization and partial ordering,
It is characterized in that, the distributed characterization vector of each document and word term pass through PV model training and obtain.
5. the distributed information retrieval based on distributed characterization and partial ordering as described in any one of Claims 1 to 4
Set option method, which is characterized in that the sample set of each set is by taking a group photo default sample rate using based on inquiry to the collection
The method of sampling sample to obtain.
6. the distributed information retrieval set option method as claimed in claim 5 based on distributed characterization and partial ordering,
It is characterized in that, the scoring of step 3 set of computations c includes:
Step 300, it determines that the document for meeting following condition is used as most relevant documentation from the sample set of the set, and forms most phase
Close document sets:
Wherein, dlFor in sample set score descending sort when ranking be l document,For sample dlScoring, τcFor for
The preset scoring threshold value of set c, ωcFor for the preset scoring rank threshold of set c；
Step 301, the scoring R of set of computations c according to the following formulac:
Wherein, αcSample rate when to be sampled to set c, πcFor the most set of relevant documents of set c.
7. the distributed information retrieval set option method as claimed in claim 6 based on distributed characterization and partial ordering,
It is characterized in that, τcIt sets according to the following formula:
τc=β × Dd,
Wherein, DdFor the scoring for the highest document d that scores in sample set, β is global parameter, and value range is [0,1].
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610251677.1A CN105956010B (en) | 2016-04-20 | 2016-04-20 | Distributed information retrieval set option method based on distributed characterization and partial ordering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610251677.1A CN105956010B (en) | 2016-04-20 | 2016-04-20 | Distributed information retrieval set option method based on distributed characterization and partial ordering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105956010A CN105956010A (en) | 2016-09-21 |
| CN105956010B true CN105956010B (en) | 2019-03-26 |
Family
ID=56917876
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610251677.1A Expired - Fee Related CN105956010B (en) | 2016-04-20 | 2016-04-20 | Distributed information retrieval set option method based on distributed characterization and partial ordering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105956010B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10572221B2 (en) | 2016-10-20 | 2020-02-25 | Cortical.Io Ag | Methods and systems for identifying a level of similarity between a plurality of data representations |
| CN107133274B (en) * | 2017-04-10 | 2020-12-15 | 浙江鸿程计算机系统有限公司 | Distributed information retrieval set selection method based on graph knowledge base |
| CN107229738B (en) * | 2017-06-18 | 2020-04-03 | 杭州电子科技大学 | Academic paper search ordering method based on document scoring model and relevancy |
| US11734332B2 (en) | 2020-11-19 | 2023-08-22 | Cortical.Io Ag | Methods and systems for reuse of data item fingerprints in generation of semantic maps |
| CN113709125A (en) * | 2021-08-18 | 2021-11-26 | 北京明略昭辉科技有限公司 | Method and device for determining abnormal flow, storage medium and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU1907300A (en) * | 1998-11-30 | 2000-06-19 | Justsystem Corporation | Term-length term-frequency method for measuring document similarity and classifying text |
| CN101582085A (en) * | 2008-09-19 | 2009-11-18 | 江苏大学 | Set option method based on distributed information retrieval system |
| CN102521350A (en) * | 2011-12-12 | 2012-06-27 | 浙江大学 | Selection method of distributed information retrieval sets based on historical click data |
| CN104050235A (en) * | 2014-03-27 | 2014-09-17 | 浙江大学 | Distributed information retrieval method based on set selection |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7287025B2 (en) * | 2003-02-12 | 2007-10-23 | Microsoft Corporation | Systems and methods for query expansion |
-
2016
- 2016-04-20 CN CN201610251677.1A patent/CN105956010B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU1907300A (en) * | 1998-11-30 | 2000-06-19 | Justsystem Corporation | Term-length term-frequency method for measuring document similarity and classifying text |
| CN101582085A (en) * | 2008-09-19 | 2009-11-18 | 江苏大学 | Set option method based on distributed information retrieval system |
| CN102521350A (en) * | 2011-12-12 | 2012-06-27 | 浙江大学 | Selection method of distributed information retrieval sets based on historical click data |
| CN104050235A (en) * | 2014-03-27 | 2014-09-17 | 浙江大学 | Distributed information retrieval method based on set selection |
Non-Patent Citations (1)
| Title |
|---|
| Collection selection for managed distributed document databases;Daryl D"Souza 等;《Information Processing and Management》;20040531;第40卷(第3期);527-546 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105956010A (en) | 2016-09-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105956010B (en) | Distributed information retrieval set option method based on distributed characterization and partial ordering | |
| CN110442777B (en) | BERT-based pseudo-correlation feedback model information retrieval method and system | |
| El-Beltagy et al. | KP-Miner: A keyphrase extraction system for English and Arabic documents | |
| CN102253982B (en) | Query suggestion method based on query semantics and click-through data | |
| US9106698B2 (en) | Method and server for intelligent categorization of bookmarks | |
| US9009146B1 (en) | Ranking search results based on similar queries | |
| US8983971B2 (en) | Method, apparatus, and system for mobile search | |
| KR101775061B1 (en) | Systems and methods for identifying aspects associated with entities | |
| Froud et al. | Arabic text summarization based on latent semantic analysis to enhance arabic documents clustering | |
| CN109960756B (en) | News event information induction method | |
| Nabli et al. | Efficient cloud service discovery approach based on LDA topic modeling | |
| CN106815252A (en) | A kind of searching method and equipment | |
| CN103838833A (en) | Full-text retrieval system based on semantic analysis of relevant words | |
| CN111104488B (en) | Method, device and storage medium for integrating retrieval and similarity analysis | |
| Froud et al. | Stemming and similarity measures for Arabic Documents Clustering | |
| CN108182186B (en) | Webpage sorting method based on random forest algorithm | |
| CN113821646A (en) | Intelligent patent similarity searching method and device based on semantic retrieval | |
| KR101377447B1 (en) | Multi-document summarization method and system using semmantic analysis between tegs | |
| US7747618B2 (en) | Augmenting user, query, and document triplets using singular value decomposition | |
| US9465875B2 (en) | Searching based on an identifier of a searcher | |
| CN108509449B (en) | Information processing method and server | |
| CN111737413A (en) | Feedback model information retrieval method, system and medium based on concept net semantics | |
| Šimko et al. | Semantic history map: Graphs aiding web revisitation support | |
| CN116450772A (en) | Intelligent recommendation method and device for search results and unified search method | |
| Wang et al. | CMU OAQA at TREC 2015 LiveQA: Discovering the Right Answer with Clues. |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190326 |